Adapted by:
Javier Pastor
Senior WriterComputer scientist turned tech journalist. I've written about almost everything related to technology, but I specialize in hardware, operating systems and cryptocurrencies. I like writing about tech so much that I do it both for Xataka and Incognitosis, my personal blog.
Karen Alfaro
WriterCommunications professional with a decade of experience as a copywriter, proofreader, and editor. As a travel and science journalist, I've collaborated with several print and digital outlets around the world. I'm passionate about culture, music, food, history, and innovative technologies.
A group of researchers has proposed a bold idea: the democratization of meteorology. Their new weather forecasting system challenges traditional models, which demand substantial computational resources. By leveraging AI algorithms, this system allows (almost) anyone to become a meteorologist and generate personalized predictions.
Aardvark Weather. According to its developers, this new weather forecasting system can turn any researcher with a desktop PC into a full-fledged meteorologist. The system uses AI algorithms as an alternative to conventional models that require extensive computing power.
(Almost) homemade forecasts. Traditional weather forecasting platforms typically take several hours to generate a forecast. They also require supercomputers and expert teams to develop, maintain, and deploy these systems. Aardvark Weather enables users to train an AI model with data from weather stations, satellites, ships, and airplanes worldwide, then generate predictions based on that data.
The research. Researchers from the University of Cambridge, the Alan Turing Institute, Microsoft Research, and the European Centre for Medium-Range Weather Forecasts (ECMWF) conducted the study, which was published this week in Nature. The researchers explain how machine learning and neural networks are replacing numerical weather prediction, enhancing both the speed and accuracy of forecasts.
Hyper-local forecasts. The system could provide hyper-localized forecasts tailored to specific industries. Richard Turner, a professor of machine learning at the University of Cambridge, told The Guardian that this model could help predict temperatures for agricultural crops in parts of Africa or wind speeds for a renewable energy company in Europe.
Weather for the next eight days. Turner adds that the model could provide accurate forecasts up to eight days in advance. In contrast, most current forecasts are only reliable for up to five days.
Lightning fast. This system can generate a complete forecast from observational data in just one second by processing it on four Nvidia A100 GPUs. By comparison, the ECMWF HRES model requires 1,000 node hours to perform the same task.
Ideal for developing countries. In regions where accurate forecasts are particularly crucial, a “tailor-made” system would be highly beneficial. According to its creators, Aardvark Weather offers this capability due to its accessibility in both implementation and use.
Previous attempts. In late 2023, DeepMind announced GraphCast, an AI-based weather forecasting system up to 1,000 times more energy-efficient than conventional models. Its accuracy even surpassed the best existing systems, yet it hasn’t been widely implemented. A few months ago, DeepMind researchers introduced Gen Cast, its successor, another machine learning-based prediction system that improves on GraphCast and competes with Aardvark Weather. AI-driven forecasting models are gaining momentum and interest, but experts have yet to determine whether they’ll be widely adopted.
Image | Brian McGowan (Unsplash)
Related | The Best Weather Apps for Your Phone
Log in to leave a comment